Date: Thu, 24 Jul 2003 01:51:46 -0700
Reply-To: Tomasz <zombek_1@YAHOO.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Tomasz <zombek_1@YAHOO.COM>
Subject: Re: Reason for transforming variables
Content-Type: text/plain; charset=ISO-8859-1
Thank you for your suggestions. I will try to find myself whether
transformations are crucial in analysing my data. Let's hope this will
improve the performance of my models
"Arthur Tabachneck" <firstname.lastname@example.org> wrote in message news:<GCETa.23874$Ii1.email@example.com>...
> Since normality is an assumption of many statistics, and many variables
> don't normally have normal distributions (but would if transformed e.g.,
> using a square root or log transformation), SAS provides a means for
> performing such transformations on the fly.
> The key question is whether your data resemble distributions for which such
> transformations are applicable (e.g., skewed or Poisson).
> "Tomasz" <firstname.lastname@example.org> wrote in message
> > Dear SAS-L,
> > Working with Enterprise Miner I have found a Variable Transform Node.
> > Due to the lack of information on the internet I am posting my
> > question on this forum.
> > What is the reason for transforming variables into normally
> > distributed?
> > I think this will affect the results of my analysis. Is it better to
> > transform or not. I could not find any paper saying clearly about the
> > raesons for transformations, only statments: "it is commonly used..."
> > But I wonder why.
> > If any of you faced this kind of problem, please let me know your
> > suggestions.
> > Tomasz